Do amplifiers really sound the same?

[James]

The analogy with photography fails because the photograph is analogous with the recording, not the reproduction

The analogy is between lens and amplifier, not between photography and hifi reproduction or the recording industry. It's one component in their respective chain that is the subject of my transparency monologue. Would the use of "optics" instead of "photography" reduce the confusion? What if I inserted the amplifier in the recording chain, would the analogy be more accurate then?

In the same way, one can photocopy a photograph and assess the accuracy of reproduction using suitable optical instrumentation.

SImilarly, one can take the original photograph and compare it visually with the photocopy. Can they be told apart? If original photographs and photocopies are shuffled, can they be identified visually as to which is which? If not, with any statistical accuracy, then the photocopier is, for those observers, transparent.

That's a good example, but would need the print media to be identical.

Similarly, if a SWBT fails to identify the path with the PAuT, then the PAuT must be transparent for those listeners.

Are you suggesting amplifier transparency is relative and subjective?

 

[Werner]

A photocopier is analogous with HiFi reproduction.

To be more precise: with the first part of the reproduction chain,
but excluding the final rendition for presentation to the user's ears.

Similarly, if a SWBT fails to identify the path with the PAuT, then the PAuT must be transparent for those listeners.

Under the circumstances of the test, and for the test stimuli as used. It is formally impossible to make valid conclusions for other circumstances and other stimuli, although with a well-executed test reasonable assumptions can be made.

 

[sergeauckland]

The analogy is between lens and amplifier, not between photography and hifi reproduction or the recording industry. It's one component in their respective chain that is the subject of my transparency monologue. Would the use of "optics" instead of "photography" reduce the confusion? What if I inserted the amplifier in the recording chain, would the analogy be more accurate then?


That's a good example, but would need the print media to be identical.


Are you suggesting amplifier transparency is relative and subjective?

Yes, that was given as oitherwise just touching the photo/photocopy would identify which is which.

Amplifier transparency is a statistical process, as it is assessed on a yes/no basis by a listener, or group of listeners. One person's hearing might be a lot more sensitive than another's, so that their threshold for transparency would be greater than someone with less good hearing.

However, over the years, certain criteria have emerged that appear still to hold true for the (great?) majority of the population.

As I've previously said, hearing acuity is a continuum, but so many on here claim to be at the very pinnacle of acuity as they seem to be able to hear things that defy ordinary mortals, like cable differences, capacitor differences, even the stand their amplifier sits on....

S.

 

[GTM]

The analogy is between lens and amplifier...

And it's a perfectly valid analogy to my mind. Both have the job of taking an input and transferring it to the output with the lowest levels of possible distortions. In that sense they perform identical functions. The fact that one transfers signals that are (representations) of the audio band and the other signals in the visible light spectrum is moot. The only difference between them is that an amplifier seeks to alter the absolute level of those signals as they pass through, a lens obviously doesn't. So it could be argued that actually a lens seeks to be transparent in the true sense of the word and an amplifier doesn't as it seeks to "distort" the input signal only it aims to do so in a wholly linear way.

 

[ultrawomble]

A lens inverts the signal (image) - does an amplifier?

 

[James]

The only difference between them is that an amplifier seeks to alter the absolute level of those signals as they pass through, a lens obviously doesn't.

This is partly true. An amplifier obviously amplifies its signal, whereas a lens can only attenuate (stopping down its aperture). A good lens should be equally sharp/accurate at all apertures, but few are.

 

[James]

A lens inverts the signal (image) - does an amplifier?

Some do.

 

[spxy]

Amplifier transparency is a statistical process, as it is assessed on a yes/no basis by a listener, or group of listeners. One person's hearing might be a lot more sensitive than another's, so that their threshold for transparency would be greater than someone with less good hearing.


S.

I dont think its as simple as some people having "better" hearing.

We learn how to perceive the world and articulate differences our scenes pick up. And this goes on through life.

We all realise we must practise in order to get better at some thing, and its the same with out scenes.

I could show someone a 3D generated image with and without various physics based illumination techniques, if they don't know what they are looking for they may say they are the same.

But once noticed and if it matters to you, you cant help but notice every time.

Because many of these blind tests seek a consensus rather than seeking to find if their are real differences i'm not sure how many hold true.

 

[ultrawomble]

Because many of these blind test seek a consensus rather than seeking to find if their are real differences i'm not sure how many hold true.

But what if a there is a consensus that a real difference exists when doing a blind test? Is that a valid result?

 

[sq225917]

I'm with SPXY here, (he says putting his architectural radiosity solution to one side).

The ABX assumes that we know what differences to listen out for. There may be differences, but we may not actively be aware of them and as such we miss them.

Assume you are presented with two images on screen, exactly the same apart from one pixel. Chances are you won't see that pixel, but once it has been pointed out to you, you will use its presence to identify the two images each and every time.

Listening to musical extracts is like being shown picture flash cards, you can't necessarily take it all in immediately. You may become reasonably adept at identifying gross differences quite quickly, as with the flash cards, but only greater long term exposure will help you develop the level of insight required to discern the smallest differences in musical presentation.

I'm all for blind tests, there really is no other way, but I do think they should be extended blind tests. 5 seconds of ABX tells you precious little at the fine end of the scale.

 

[sergeauckland]

But what if a there is a consensus that a real difference exists when doing a blind test? Is that a valid result?

I would say so yes.

If several people identify a difference, and when the results are analysed, they are right rather better than just guessing, then that's definitely a valid result.

The more people that take the test, then the better the statistics are in terms of accuracy of result.One peson taking the test once could be a guess. One person taking the test 10x and getting it right could be a very lucky guess. One person taking the test 100x, and the changes of guessing reduces greatly. Multiply that now by several people taking the test, and the accuracy of the prediction improves. There are lots of statistical methods to eliminate lucky guesses, or deliberate attenpts to thwart the test.

S.

 

[sergeauckland]

I'm with SPXY here, (he says putting his architectural radiosity solution to one side).

The ABX assumes that we know what differences to listen out for. There may be differences, but we may not actively be aware of them and as such we miss them.

Assume you are presented with two images on screen, exactly the same apart from one pixel. Chances are you won't see that pixel, but once it has been pointed out to you, you will use its presence to identify the two images each and every time.

Listening to musical extracts is like being shown picture flash cards, you can't necessarily take it all in immediately. You may become reasonably adept at identifying gross differences quite quickly, as with the flash cards, but only greater long term exposure will help you develop the level of insight required to discern the smallest differences in musical presentation.

I'm all for blind tests, there really is no other way, but I do think they should be extended blind tests. 5 seconds of ABX tells you precious little at the fine end of the scale.

AB or ABX testing doesn't have to be 5 seconds, or any specific time, if the subject undergoing the test has control of the switching. Some things are very clear with rapid AB switching, others are better judged over a longer period. For example, rapid AB switching shows up level differences very clearly, which is why level matching has to be done to a high level, say 0.1dB if not better.

S.

 

[hecd2]

I agree that if transfer functions are identical, then there should be no difference in either amplifiers or lenses. I can certainly hear no difference between my "lesser" Densen B-340 4-channel amplifier when compared with the "better" B-350 mono-blocks; nor can I discern any image difference between my Pentax-A 50/1.4 and FA50/1.4, which have identical optics.

So we agree that it's possible to make a measurement both with amplifiers and lenses which, if the same within the perceptible region, will characterise performance that is indistinguishable. The question of defining the perceptible region then arises.

But when comparing two differently designed, but similarly specified, items - their measured performance may show a difference (if measuring the right things) or it may not.

Indeed, and it is the case that if they do not show a difference then their performance will be indistinguishable.

Serge's contention that most SS amplifiers are practically transparent and therefore should sound the same may not hold if he's not measuring, for example, their transfer function. In any case, one of his criterion for transparency is a flat response from 20Hz - 20kHz, but does this mean an ultra-wide bandwidth design that is flat from DC to 2MHz, but identical in every other measureable aspect to another that has a brickwall filter at 22kHz, should sound the same? I'm not sure.

Well, the amplitude and phase frequency response is the Fourier transform of the point spread function or impulse response - in other words it is characterised by the transfer function and the output is determined by convolving the PSF with the input. As I said, it then becomes of a question of determining the perceptible region. I suggest that measuring the frequency response of an audio system at 2MHz is as pointless as measuring the performance of an optical lens at X-ray or radio frequencies. I'm not sure why any audio amplifier would have a brickwall filter at 22kHz, but even if it did I don't see how the output would be distinguishable at frequencies from one that didn't have it provided it doesn't affect the frequency response (in optical terms the modulation transfer function) at audible frequencies - ie the rolloff shouldn't start until above 20kHz or whatever we take as the limit of audible perception (any more than the fact that the UV and IR filtering inherent in photography affects the performance of the system in the visible).

When I changed from Naim NAP135s to Densen B-350s, I was struck by a very distinct change in sonic character that I could pick them apart blind. It's probable the Naim has a different transfer function to the Densen, but both very competent amplifiers can't be considered transparent if they sounded different.

Well they both can't be transparent if they are perceptibly different - but perhaps one is transparent (ie flat frequency response in the audible region) and the other isn't.

 

[James]

Listening to musical extracts is like being shown picture flash cards, you can't necessarily take it all in immediately. You may become reasonably adept at identifying gross differences quite quickly, as with the flash cards, but only greater long term exposure will help you develop the level of insight required to discern the smallest differences in musical presentation.

Some can stare for a long time and still not see any differences.

 

[Werner]

The more people that take the test, then the better the statistics are

Yes. But not just any people.

Imagine two components that are seemingly indentical, except for a subtle difference in their presentation of the stereo image.

When presentated to non-enthusiasts this may well go unnoticed, as it is suspected that people not seriously engaged in audio/music hardly perceive stereo imaging.

Even when confining the test to experienced listeners this still may happen. I myself am mostly deaf to imaging when presented with an alien system and/or an alien room. In some cases it takes me several hourse of listening before I am able to decode what the speakers are hurtling at me.


So ideally difference testing is done on large populations, but the subjects must be of sufficient training, and they must be in a sympathetic environment.

I take blind testing results stemming from large amounts of people dumped in arbitrary rooms with arbitrary systems and arbitrary music with a large pinch of salt.

And so the ultimate ABX test appears to be rather impractical.

 

[ultrawomble]

Some can stare for a long time and still not see any differences.

Squad cars .......

 

[Robert]

The ABX assumes that we know what differences to listen out for. There may be differences, but we may not actively be aware of them and as such we miss them.

Which surely goes back to a point I made the other day that we should be looking at the relevance of any difference shown to exist as by far the most important consideration.

I'm not arguing that you rely solely on a quick 10 second ABX, though even that is far preferable to any sighted comparison which is seriously flawed from the start, but if you repeat the process many times and still cannot detect any difference, any that does exist is of no significance.

I don't however accept that such a condition can occur - where a difference is heard when a comprehensive bench test shows all distortion to be extremely low. You can argue the toss if you've got 0.2% of primarily crossover related distortion, or 3% low order THD from a tube amp but nearly all competent electronics produced over the past few decades give such low figures, and perform well enough is all other areas that comparisons become moot.
Any difference found under such circumstances will be caused by interface issues, which are of course understood and no mystery surrounds them.

Some can stare for a long time and still not see any differences.

This makes my point perfectly:

There is a difference but it doesn't alter the viewer's reaction to the photograph because it isn't material. However if you made the sky pink, altered the aspect ratio or inserted a few horse horse drawn carriages..... (in other words, the equivalent of audible sonic differences)

 

[ultrawomble]

a large pinch of salt.

Would that pinch of salt be larger or smaller than one you take with say press reviews or brochures?

 

[hecd2]

Some can stare for a long time and still not see any differences.

That's nothing like optical performance or audio presentation so now the analogy is getting completely strained. What we have here are two very busy images which are identical apart from a number of manipulated details. It would be like trying to distinguish between the same recording played twice on the same equipment except that in one case a few single notes were edited out or changed. In both cases, the photos above and the hypothetical recording, the differences are objectively present and perceptible.

 

[Mescalito]

But what if a there is a consensus that a real difference exists when doing a blind test? Is that a valid result?

Yes, it is.

But measuring the amps in question would reveal the cause of the difference.

Chris